[ofa-general] Re: [PATCHv3][TRIVIAL] opensm/osm_sa_mcmember_record.c: Some error message improvements

Sasha Khapyorsky sashak at voltaire.com
Mon Jun 16 08:16:03 PDT 2008


On 06:11 Mon 16 Jun     , Hal Rosenstock wrote:
> > > @@ -1278,8 +1281,11 @@ __osm_mcmr_rcv_leave_mgrp(IN osm_sa_t * sa,
> > >  		}
> > >  	} else {
> > >  		CL_PLOCK_RELEASE(sa->p_lock);
> > > -		OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
> > > -			"Failed since multicast group not present\n");
> > > +		OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 1B31: "
> > > +			"Failed since multicast group 0x%16"
> > 
> > Why to change log level? It is not OpenSM error, but invalid request.
> 
> Are errors really restricted to OpenSM (internal) errors ?

I guess it should be.

> I thought
> they were more inclusive than that including network events that were
> "interesting events".

We have VERBOSE level for "interesting" events and INFO for "very
interesting" ones.

> In this specific case, I agree it is a bad idea as I think restarting
> IPoIB does causes this code path to be used. Modified patch already
> issued.

Cool. Thanks.

Sasha



More information about the general mailing list